摘要
不断提高合成孔径雷达(SAR)图像目标识别能力对于全天时、全天候战场情报侦察具有重要意义。近年来,深度学习模型在SAR目标识别领域得到了广泛应用和验证,但由于SAR图像样本往往十分有限,模型的适应性受到一定制约。提出结合三维电磁散射模型和深度学习的SAR目标识别框架,充分运用三维电磁散射模型在目标SAR数据生成以及物理属性描述方面的优势,提升深度学习模型的分类可靠性和适应性。
It is of important meaning to consistently improve Synthetic Aperture Radar(SAR)target recog-nition capability for all-day,all-weather battlefield reconnaissance.In recent years,deep leaning models have widely used and verified in the field of SAR target recognition.However,as the samples of SAR images are often very limited,the adaptivity of the models is restricted to some extent.A framework for SAR target recognition is proposed via the combination of 3-D scattering center model and deep learning,which comprehensively employs the advantages of 3-D scattering center model for SAR data generation and physical properties descriptions thus improving the classification reliability and adaptivity of deep learning models.
作者
丁柏圆
周春雨
Ding Baiyuan;Zhou Chunyu(Unit 96901 of PLA,Beijing 100094,China)
出处
《航天电子对抗》
2024年第2期34-38,64,共6页
Aerospace Electronic Warfare
基金
国家自然科学基金(62001501)。
作者简介
丁柏圆(1990-),男,助理研究员,博士。
